Welcome to the Dendr AI Blog

Your AI agents are already taking actions no one explicitly authorized. This blog is where we work through what it takes to fix that.

Enterprises are wiring together more and more agents — one reads a report, another drafts a message, another calls a third-party tool. Each step looks reasonable in isolation, but nobody signed off on the chain as a whole, and nobody can prove what happened after the fact. That gap between what agents actually do and what anyone can verify is the problem Dendr AI exists to close.

We started Dendr AI around two ideas that most orchestration tools don't hold together: coordination paths should be computed dynamically from a goal instead of hand-authored in advance, and authorization should cover the entire chain at once instead of gating every checkpoint along the way. Here, we'll write about how that plays out in practice — the authorization model, the conformance testing we run on every registered agent, and the lessons we pick up along the way from CISOs, auditors, and platform teams putting this into production.

More posts are on the way. If you want the short version in the meantime, the homepage lays out how the whole system fits together.